why UDF have so ugly filesize limit?

why UDF have so ugly filesize limit?

[Marat Buharov]

from fs/udf/super.c:
in function udf_fill_super
sb->s_maxbytes = 1<<30; (1 GB)

Why sb->s_maxbytes is not equal to MAX_LFS_FILESIZE?

So, in include/linux/fs.h written that the filesystems should put that
(MAX_LFS_FILESIZE) into their s_maxbytes, otherwise bad things can
happen in VM.

Re: why UDF have so ugly filesize limit?

[Chuck Ebbert]

Marat Buharov wrote:
> from fs/udf/super.c:
> in function udf_fill_super
> sb->s_maxbytes = 1<<30; (1 GB)
> 
> Why sb->s_maxbytes is not equal to MAX_LFS_FILESIZE?
> 

Patches to fix that are in the -mm kernel already (and in
Fedora Core 6 latest update.)

Re: why UDF have so ugly filesize limit?

[Alan Cox]

On Thu, 19 Apr 2007 12:56:27 -0400
Chuck Ebbert <cebbert@redhat.com> wrote:

> Marat Buharov wrote:
> > from fs/udf/super.c:
> > in function udf_fill_super
> > sb->s_maxbytes = 1<<30; (1 GB)
> > 
> > Why sb->s_maxbytes is not equal to MAX_LFS_FILESIZE?
> > 
> 
> Patches to fix that are in the -mm kernel already (and in
> Fedora Core 6 latest update.)

Good to know, presumably these also fix the underlying UDF problems in the
UDF filesystem code that caused it to be put there in the first pace ?

(Security CVE-2006-4145)

Alan

Re: why UDF have so ugly filesize limit?

[Chuck Ebbert]

Alan Cox wrote:
> On Thu, 19 Apr 2007 12:56:27 -0400
> Chuck Ebbert <cebbert@redhat.com> wrote:
> 
>> Marat Buharov wrote:
>>> from fs/udf/super.c:
>>> in function udf_fill_super
>>> sb->s_maxbytes = 1<<30; (1 GB)
>>>
>>> Why sb->s_maxbytes is not equal to MAX_LFS_FILESIZE?
>>>
>> Patches to fix that are in the -mm kernel already (and in
>> Fedora Core 6 latest update.)
> 
> Good to know, presumably these also fix the underlying UDF problems in the
> UDF filesystem code that caused it to be put there in the first pace ?
> 
> (Security CVE-2006-4145)

Yeah, the patches are by Jan Kara who put the original limitation
in there as a temporary band-aid.

Re: why UDF have so ugly filesize limit?

[Jan Kara]

> from fs/udf/super.c:
> in function udf_fill_super
> sb->s_maxbytes = 1<<30; (1 GB)
> 
> Why sb->s_maxbytes is not equal to MAX_LFS_FILESIZE?
  Because UDF had some flaws and user could crash a kernel with larger
filesize. In -mm kernel are patches fixing the flaw and also raising the
limit back to MAX_LFS_FILESIZE.

> So, in include/linux/fs.h written that the filesystems should put that
> (MAX_LFS_FILESIZE) into their s_maxbytes, otherwise bad things can
> happen in VM.
  Bad things can happen only if you set it to more than
MAX_LFS_FILESIZE. With smaller values only users are disappointed ;).

								Honza
-- 
Jan Kara <jack@suse.cz>
SuSE CR Labs